兔子API 兔子API
中

model

OpenAI GPT and o-series

Direct answerAs reviewed on 2026-08-08, OpenAI positions GPT-5.6 Sol, Terra, and Luna for complex professional work, balanced intelligence and cost, and high-volume workloads through the Responses API.

Updated · Reviewed

Current official line

As of this review, OpenAI's catalog recommends GPT-5.6 Sol for complex reasoning and coding, Terra for a balance of intelligence and cost, and Luna for cost-sensitive high-volume work. Exact API IDs, aliases, and snapshots remain governed by the live official catalog.

Modalities and tools

OpenAI documents text and image input, text output, multilingual capability, and vision across the current GPT-5.6 line and lists functions, Web Search, File Search, and Computer Use. Availability also depends on account, channel, and API route.

Choosing within the family

Evaluate Sol for long-running engineering, dense constraints, and high-value decisions; start with Terra for routine agents, coding, and enterprise content; consider Luna for classification, extraction, simple generation, and high concurrency. Compare quality, latency, and total tokens on the same real workload.

API use

Responses is the primary official interface. If the exact ID is enabled here, call /v1/responses; Chat Completions compatibility must be confirmed in the marketplace and protocol tests. Image-generation models are intentionally excluded from the gpt- text-family match.

Versions and limits

Aliases can move, while pinned snapshots improve reproducibility. Context and output limits are not a promise that both can be maximized together; verify tools, structured output, rate limits, geography, data terms, and retirement dates.

Use cases

  • General reasoning
  • Complex coding
  • Vision and tool use

API protocols

  • /v1/chat/completions

FAQ

Is the model name permanent?

No. Official aliases can move and enabled API IDs can change; pin a tested ID in production and verify availability separately in the model marketplace.

How should I choose among capability-first, balanced, and high-throughput GPT models?

Filter by task value and difficulty, then compare accuracy, tool success, first-token and total latency, and total tokens on the same real samples. Simple classification or extraction need not default to the highest-capability tier; evaluate it for complex engineering and dense constraints.

Can every GPT model use both Responses and Chat Completions?

Not necessarily. Responses is the primary interface for the current official line, but model and endpoint availability must be checked separately on this site or channel. Chat compatibility also needs protocol tests; a gpt- prefix does not guarantee identical tools or fields.

  • OpenAIOpenAI recommends Responses API for new projects. Direct and gateway calls both use a Bearer header, but require separate OpenAI and site keys, and the gateway endpoint must be implemented for the selected model.
  • OpenAI-compatible API guideThis site's OpenAI-compatible routes use a Bearer token; Responses, Chat Completions, and media endpoints have distinct objects and streaming semantics and must match the target model.
  • Choosing and calling coding modelsSelect coding models by patch correctness, test pass rate, tool success, latency, and cost on real repositories, not marketplace copy or a single benchmark.
  • Tool use and AI agent model guideAn agent is a controlled loop, not merely tool calling: structured schemas, least privilege, state, idempotency, timeouts, approval, audit, and replayable evaluation.

Official sources

  1. OpenAI Models Official